DMWM: Dual-Mind World Model with Long-Term Imagination

arXiv 25.2 2025 59.6 method

TLDR

Proposes DMWM, a dual-mind world model integrating logical reasoning with RSSM for long-term imagination, evaluated on DMControl.

Reasoning

The paper introduces a novel dual-process framework combining intuitive RSSM and logical reasoning to address long-term prediction errors, which is a clear strength. However, evaluation is limited to simulated DMControl benchmarks without real-world experiments, and the abstract lacks details on scalability or comparison to other approaches.

Read-first score

Read-first score 59.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 46.

Recency 8%

86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=benchmark,experiment,result

Topical relevance 42%

65.7

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 267.

Keyword Scores

world model

model-based reinforcement learning world model

world dynamics prediction

generative world model

world simulator

interactive world model

video world model

Deep Analysis

Innovations

Dual-mind architecture combining RSSM-based System 1 (intuitive) and logic-integrated neural network System 2 (logical reasoning) for world models
Inter-system feedback mechanism to ensure imagination follows logical rules of the real environment
Hierarchical deep logical reasoning in LINN-S2 to guide long-term imagination with logical consistency

Methodology

DMWM consists of two components: an RSSM-based System 1 (RSSM-S1) for intuitive state transitions and a logic-integrated neural network-based System 2 (LINN-S2) for hierarchical deep logical reasoning. An inter-system feedback mechanism ensures the imagination process adheres to logical rules. The framework is evaluated on long-term planning tasks from the DMControl suite.

Key Results

DMWM yields significant improvements over state-of-the-art world models in logical coherence, trial efficiency, data efficiency, and long-term imagination on benchmark tasks requiring long-term planning.

Resources

Paper